Please take our brief survey

Blueprints Programs = Positive Youth Development

Return to Search Results

Model Program Seal

Maryland Ignition Interlock License Restriction (Policy)

Blueprints Program Rating: Model

A statewide license restriction program for drivers with multiple alcohol-related traffic offenses to reduce alcohol-impaired driving recidivism.

  • Department of Transportation National Highway Traffic Safety Administration
  • 1200 New Jersey Avenue SE
  • Washington, DC 20590
  • www.nhtsa.gov/
  • Adult Crime

    Program Type

    • Adult Crime Prevention
    • Police Crime Prevention

    Program Setting

    • Community (e.g., religious, recreation)

    Continuum of Intervention

    • Indicated Prevention (Early Symptoms of Problem)

    A statewide license restriction program for drivers with multiple alcohol-related traffic offenses to reduce alcohol-impaired driving recidivism.

      Population Demographics

      Adult drivers with multiple alcohol-related driving offenses whose licenses have been revoked or suspended and who were later approved for a restricted license to drive an interlock-equipped vehicle.

      Age

      • Adult

      Gender

      • Male and Female

      Race/Ethnicity

      • All Race/Ethnicity

      Race/Ethnicity/Gender Details

      Voas et al. (2016) found no differences between groups in outcomes by gender or ethnicity.

      Opportunity to drive or start a vehicle while alcohol impaired

      • Individual

      The use of ignition interlock license restrictions requires those with previous alcohol-related driving offenses to pass a breath analyzer test of blood alcohol levels before starting a vehicle. The program further requires that drivers who are applying for reinstatement of their driving licenses install the ignition interlock system in their vehicle and continue to use it for two years before they are eligible for reinstatement.

      Note: The State of Maryland evaluated the effects of interlock installation in two studies. Both were RCTs and carefully conducted. One involved one year of installation and one study two years. Both found large reductions during installation. The one-year study did not show sustained effects, while the two year study did. Thus, Blueprints is certifying only the two-year study.

      The use of ignition interlock license restrictions requires those with previous alcohol-related driving offenses to pass a breath analyzer test of blood alcohol levels before starting a vehicle. By using this automated system, intersecting risk behaviors (drinking and driving) are controlled, rather than either behavior separately. The program further mandates that drivers who are applying for reinstatement of their driving licenses install the ignition interlock system in their vehicle and continue to use it for two years before they are eligible for reinstatement.

      The ignition interlock device requires a potential driver to blow into a breath alcohol sensor connected to the vehicle's ignition system before the vehicle's engine will start. An on-board computer analyzes the alcohol concentration of the driver's breath and compares it to a set point, usually .02 grams per deciliter. If the alcohol concentration is above the set point or the driver does not provide a breath sample, the interlock prevents the vehicle engine from starting.

      Ignition interlocks are comprised of four basic elements:

      1. A breath alcohol sensor installed in the passenger compartment of a vehicle and connected to a control unit in the engine compartment that allows the engine to start only upon an acceptable breath test;
      2. A tamper-proof system for mounting the control unit in the engine compartment;
      3. A data-recording system that logs breath test results, tests compliance, and collects other state-mandated data; and
      4. A retest system which, after the engine has started, requires the driver to provide additional breath samples (generally every 10 to 15 minutes) to ensure that the driver remains alcohol-free. Several minutes are provided for the driver to exit traffic and move to a safe location for retesting.

      By addressing the combined behaviors of driving and drinking, the ignition interlock system bypasses disincentives that are presumed to motivate the alcohol-intoxicated driver and reduce unsafe driving.

      • Behavioral

      The two qualifying studies for Blueprints were Beck et al. (1999) and Rauch et al. (2011). Beck et al. (1999) randomly assigned 1,387 individuals who had previously lost their driving licenses due to alcohol-related offenses and were applying to have their licenses reinstated to either a control group (n = 689) or the treatment group (n = 698). Those in the treatment group were required to have the ignition interlock system installed in their vehicles and to maintain the system for one year. The study then followed the participants for two years to measure alcohol-related traffic violations.

      Rauch et al. (2011) followed a similar study design, randomly assigning 1,927 drivers to the treatment (n = 944) or control (n = 983) group. However, Rauch et al. (2011) required drivers to use the ignition interlock system for two years (as opposed to one in Beck et al., 1999) and monitored the outcomes for an additional two years after completion of the intervention.

      Morse and Elliott (1992) matched 273 individuals using the ignition interlock system with individuals who had had their licenses suspended due to DUI convictions. The study then analyzed the time until re-arrest over the following 30 months.

      Voas et al. (2016) used propensity score matching to compare 640 individuals mandated to an alcohol use disorder treatment program with 806 individuals not mandated to a program, all of whom had completed a sentence with the ignition interlock system and had multiple previous DUI offenses. The study then analyzed recidivism hazard up to 1.5 years after the removal of the interlock system.

      Roth et al. (2007) also used propensity score matching to compare 437 offenders who were convicted and installed interlocks for an average of 322 days to a comparison group of 12,554 offenders convicted during the same time but who did not install interlocks during an equivalent 322-day period. Rearrest data were monitored at three time points: (1) while the interlock was on the car, (2) following removal of the interlock, and (3) over a 4-year interval that combined periods 1 and 2.

      Willis et al. (2009) presented a meta-analysis and systematic review of one RCT and 13 other controlled trials.

      Beck et al. (1999) reported that drivers with the ignition interlock system were significantly less likely to recidivate in the year in which they had the system installed as compared to the control group. In the next year, the treatment and control groups did not differ significantly on alcohol-related traffic violations, but first-year benefits were maintained. Rauch et al. (2011) found a similar positive program effect on drivers during the two years of the intervention period and a smaller, but still significant, effect in the two years after the intervention. Morse and Elliott (1992) also found a significant positive program effect: the DUI re-arrest rate for the license-suspension group was approximately three times as great as that of the interlock group. In addition, Voas et al. (2016) found that the treatment group was 32% less likely to recidivate than the comparison group. Results of Roth et al. (2007) showed a reduction in recidivism of 65% during installation for the treatment group compared to controls. In a 3-year period following removal, there was no significant difference between groups in recidivism rates. At the end of 4 years, including the period when the interlock was installed and after its removal, results showed rates were maintained, indicating that interlock users still had lower total recidivism than nonusers.

      The meta-analysis (Willis et al., 2009) found effects only when the device was installed, not afterward. The lack of long-term effects plus the low methodological quality of the studies led the authors to conclude that the program had not as yet demonstrated clear benefits.

      Drivers with one year of interlock installation, relative to controls, had (Beck et al., 1999):

      • lower risk of receiving an alcohol-related traffic violation (2.4% versus 6.7%)
      • no sustained effect in the year after the ignition interlock system was removed

      Drivers with two years of interlock installation, relative to controls, had (Rauch et al., 2011):

      • 36% lower hazard of committing an alcohol-related offense
      • sustained effects two years following the intervention period, with a 32% lower hazard rate

      In addition, Roth et al. (2007) showed that after four years, interlock users still had lower total recidivism than nonusers.

      The studies did not report formal mediator analysis.

      Beck et al. (1999) reported a small effect size in favor of treatment. That is, in the year the treatment group had the ignition interlock system in place, treatment group participants had .36 lower relative risk of an alcohol-related traffic violation than the control group. Rauch et al. (2011) also reported a small effect size favoring the treatment group. Results showed a 1.57 higher hazard of an alcohol-related offense among the control group (compared to the treatment group) in the two years of the intervention period. Morse & Elliott (1992) did not report effect sizes. Compared to controls, the average effect size for the treatment group reported in Voas et al. (2016) was medium to large, as alcohol-impaired recidivism rates for treatment were reduced by roughly one-third. Roth et al. (2007) reported a medium-large effect size at the end of a 1-year period after installation (hazard ratio = .34), and a small effect size after four years (hazard ratio = .78).

      The samples used in Beck et al. (1999), Rauch et al. (2011), Morse & Elliott (1992), Voas et al. (2016) and Roth et al. (2007) were medium-large from four different states (Maryland, Ohio, Florida and New Mexico). The meta-analysis (Willis et al., 2009) examined studies with samples from several U.S. cities and states plus samples from Canada, Australia, and Sweden.

      Beck et al. (1999)

      • The study used police records, which capture only a small portion of intoxicated driving
      • Significance tests for baseline equivalence did not include measures of past alcohol abuse and driving records
      • Outcome equals zero at baseline in survival models but no controls for past alcohol use or driving offenses

      Rauch et al. (2011)

      • The study used police records, which capture only a small portion of intoxicated driving

      Morse & Elliott (1992)

      • Used a matched comparison group design
      • Arrest records may miss instances of drunk driving
      • Significant differences between conditions existed after matching

      Voas et al. (2016)

      • Not randomly assigned to condition, matched comparison groups
      • Used police records, which may miss instances of intoxicated driving
      • Used stepwise regression to select covariates
      • Conducted an exploratory factor analysis to form a composite variable used in the propensity score match
      • Groups differed on number of violations prior to treatment (3 for treatment, 2 for control)
      • Significant differences at baseline after matching protocol

      Roth et al. (2007)

      • QED with four variables (all dichotomized) used in the propensity score model matching of controls with treatment subjects
      • Survival analysis could not control for the baseline outcome

      Willis et al. (2009)

      • The meta-analysis studies were not of high quality.

      • Blueprints: Model

      Beck, K. H., Rauch, W. J., Baker, E. A., & Williams, A. F. (1999). Effects of ignition interlock license restrictions on drivers with multiple alcohol offenses: A randomized trial in Maryland. American Journal of Public Health, 89, 1696-1700.

      Morse, B. J., & Elliott, D. S. (1992). Effects of ignition interlock devices on DUI recidivism: Findings from a longitudinal study in Hamilton County, Ohio. Crime & Delinquency, 38, 131-157.

      Rauch, W. J., Ahlin, E. M., Zador, P. L., Howard, J. M., & Duncan, G. D. (2011). Effects of administrative ignition interlock license restrictions on drivers with multiple alcohol offenses. Journal of Experimental Criminology, 7, 127-148.

      Roth, R., Voas, R., & Marques, P. (2007). Mandating interlocks for fully revoked offenders: The New Mexico experience. Traffic Injury Prevention, 8(1), 20-25.

      Voas, R. B., Tippetts, S., Bergen, G., Grosz, M., & Marques, P. (2016). Mandating treatment based on interlock performance: Evidence for effectiveness. Alcoholism: Clinical and Experimental Research, 40, 1953-1960.

      Willis, C., Lybrand, S., & Bellamy, N. (2009). Alcohol Ignition Interlock Programmes for reducing drink driving recidivism (Review). The Cochrane Library, 1, 1-26.

      Eileen Michelle Ahlin, Ph.D.
      Evaluator
      Assistant Professor
      Criminal Justice School of Public Affairs
      Penn State Harrisburg
      (717) 948-4341
      ema105@psu.edu

      Kenneth H. Beck, Ph.D., FAAHB
      Evaluator
      Professor
      Dept. of Behavioral & Community Health
      School of Public Health
      University of Maryland
      kbeck1@umd.edu

      Study 1

      Beck, K. H., Rauch, W. J., Baker, E. A., & Williams, A. F. (1999). Effects of ignition interlock license restrictions on drivers with multiple alcohol offenses: A randomized trial in Maryland. American Journal of Public Health, 89, 1696-1700.

      Study 2

      Rauch, W. J., Ahlin, E. M., Zador, P. L., Howard, J. M., & Duncan, G. D. (2011). Effects of administrative ignition interlock license restrictions on drivers with multiple alcohol offenses. Journal of Experimental Criminology, 7, 127-148.

      Beck, K. H., Rauch, W. J., Baker, E. A., & Williams, A. F. (1999). Effects of ignition interlock license restrictions on drivers with multiple alcohol offenses: A randomized trial in Maryland. American Journal of Public Health, 89,  1696-1700.

      Evaluation Methodology

      Design:

      Recruitment: The study recruited Maryland drivers with multiple alcohol traffic offenses whose licenses had been revoked or suspended and who were later approved for relicensing by the state Medical Advisory Board.

      Assignment: Participants were randomly assigned to either the business-as-usual control group (n=689) or the ignition interlock system treatment group (n=698). The control group received a probationary license agreeing not to drive after consuming any alcohol and participating in Maryland’s Drinking Driving Monitoring Program, which involves reporting to a probation monitor. The treatment group was required to install the ignition interlock device and was also informed about treatment and support programs, in which they were required to participate.

      Attrition: The subjects were followed for 1 year after the program start, until the end of the required period of use of the interlock device, and then for 1 more additional year. Of the 1,387 participants, 23 moved out of state during the study, but all were followed for the purposes of analysis.

      Sample: The sample was predominately male (88-91%) and white (82-86%). The median age was 33; participants ranged from ages 19-75.

      Measures: The study used police records on whether participants received subsequent alcohol-related traffic violations.

      Analysis: The study used relative risk analysis and proportional hazard survival models to compare the rate of recidivism in the treatment and control groups. The multivariate survival models controlled for demographic covariates but not for baseline measures of past alcohol use and driving offenses.

      Intent-to-Treat: All randomized participants were included in the analysis.

      Outcomes

      Implementation Fidelity: Of the total sample, 14% (n=98) failed to comply with the treatment requirements and remained suspended. About 57% installed the device, and an additional 23% signed a waiver stating that they did not own a vehicle and would not drive other vehicles without the ignition interlock system, and 7% signed a waiver for the restriction period, but had an interlock system installed later.

      Baseline Equivalence: All participants had multiple previous alcohol-related driving offenses. The number of previous alcohol traffic violations did not differ between the interlock group (mean = 3.57, SD = 1.43) and the control group (mean= 3.61, SD = 1.33). Table 1 shows no significant differences on the demographic measures.

      Differential Attrition: With all subjects followed, there was no attrition.

      Posttest: Within the year after assignment, those in the control group had a significantly higher risk of receiving an alcohol-related traffic violation than those in the treatment group (6.7% versus 2.4%, .36 relative risk for the treatment group).

      Long-Term: In the year after the ignition interlock systems were no longer required, there were no significant differences in the outcome measure between the treatment and control groups. Over the full 2-year period, however, the program benefits for alcohol-related traffic violations remained significant (9.1% in the control group versus 5.9% in the intervention group, and .64 relative risk for the treatment group).

      Rauch, W. J., Ahlin, E. M., Zador, P. L., Howard, J. M., & Duncan, G. D. (2011). Effects of administrative ignition interlock license restrictions on drivers with multiple alcohol offenses. Journal of Experimental Criminology, 7,  127-148.

      Evaluation Methodology

      Design:

      Recruitment: The study followed the protocol established in Study 1 by recruiting Maryland drivers who had two or more alcohol traffic offenses and who had been approved for reinstatement of their driver's license by the state’s Medical Advisory Board.

      Assignment: Participants were randomly assigned to either the control group (n=983) or the ignition interlock system treatment group (n=944). Drivers assigned to the control group reported to a probation monitor and enrolled in Maryland’s Drinking Driver Monitor Program. The treatment group was enrolled in the interlock license restriction program and had 30 days to have the interlock installed.

      Attrition: The study waited until 6 years after the start to assess subjects during the 2 years of the program and during the 2 years following the program end. Of the 1,927 participants, a total of 58 drivers died (3.0%) and 68 moved out of state during the study (3.5%).

      Sample: The mean age of the sample was 40 years at the start of the program and the sample was predominately male (88%). The sample was comprised of 80% white, 15% black, 1% Asian, and 4% other/unknown participants.

      Measures: The study used all alcohol-related violations, including any arrest for DUI or DWI that resulted in a preconviction administrative sanction, a conviction, probation before judgment, or their combination as reported by state legal records.

      Analysis: The study conducted Kaplan-Meier survival analysis to compare the proportion of the samples that did not receive alcohol-related traffic violations. The authors also used proportional hazard models to estimate the effects of potential risk factors on the probability of alcohol-related violation-free survival during the study period, including age, sex, race, prior alcohol-related violations at program enrollment, and the violation disposition category. Finally, the study estimated a hazard ratio for the outcome measure.

      Intent-to-Treat: All participants assigned to the treatment were included in the analysis. The models treated subjects who died or moved out of state as censored.

      Outcomes

      Implementation Fidelity: The treatment participants stayed in the program an average of 23 months, while the control group remained an average of 21 months. For the treatment group, 22% of the subjects failed to comply by installing the interlock ignition system or obtaining a waiver.

      Baseline Equivalence: The study reported no significant differences between treatment groups for demographic measures, the number of previous alcohol-related offenses, or the prior violation disposition category.

      Differential Attrition: There was no significant difference across conditions in the likelihood of dying or moving from the state. The study did not provide further analysis, but the attrition rate was low.

      Posttest: After controlling for age, gender, race/ethnicity, and prior offenses, the study found that in the 2-year intervention period, the treatment group had a 36% lower hazard of committing an alcohol-related offense.

      Long-Term: In the two years following the intervention period, the treatment group had 26% lower hazard of committing an alcohol-related offense and a 32% lower hazard over the entire 4-year period.

      Morse, B. J. & Elliott, D. S. (1992). Effects of ignition interlock devices on DUI recidivism: Findings from a longitudinal study in Hamilton County, Ohio. Crime & Delinquency, 38,  131-157.

      Evaluation Methodology

      Design:

      Recruitment: The study drew the sample from an eligible pool of persons convicted of a DUI in Hamilton County, Ohio, between July 1, 1987 and February 28, 1989. After conviction, judges had the option to offer the ignition interlock program or license suspension only. Offenders who were offered the ignition interlock program then had the option to participate or opt for license suspension only.

      Assignment: The study found that selection into the ignition interlock program was not random and comparing participants to those who were assigned to license suspension or those who opted for license suspension would bias results. Therefore, it matched the ignition interlock participants with a comparison group based on a cluster analysis of three measures closely related to future re-offense: 1) self-reported problem drinking classification, 2) number of prior alcohol/drug-related arrests, and 3) number of prior DUI arrests. A total of 237 matched pairs were included in the study.

      Attrition: Subjects were followed for up to 30 months. The study did not discuss attrition. Presumably, due to using police records, there was no attrition, however it is possible that participants were lost due to moving out of the area.

      Sample: The sample was predominately male (87.9%) and white (80.5%). The average age in the sample was 36.4 years and over half of the sample was categorized as working class (61.2%). A majority of the sample had at least one prior DUI arrest (67.3%).

      Measures: The measure for recidivism was re-arrest for DUI in Hamilton County for a period of up to 30 months. The records search also included driving under suspension, no driver’s license, and other alcohol and drug-related offenses. Re-arrest outside the county or DUI without arrest were not captured by the measure. Self-reported survey measures were also obtained but only at baseline.

      Analysis: The study conducted survival analysis with time-to-failure for DUI or related offenses during the time-at-risk period, which ranged from 12 to 30 months.

      Intent-to-Treat: It appears that all subjects were included.

      Outcomes

      Implementation Fidelity: The study did not report on how long participants remained in the program or how well they followed requirements for use of the ignition interlock system.

      Baseline Equivalence: After matching, some significant differences remained between conditions. Most importantly, those in the interlock condition had significantly more prior DUI arrests than license suspension members. The study reported that there were other differences, but did not consider them predictive of recidivism.

      Differential Attrition: As records for all subjects were searched, there appears to be no attrition.

      Posttest: The DUI re-arrest rate for the license suspension group was approximately three times as great as that of the interlock group across all time-at-risk periods (p=.004).

      Long-Term: With the required period for use of the ignition interlock system lasting 12 to 30 months, the study did not conduct long-term follow-up.

      Voas, R. B., Tippetts, S., Bergen, G., Grosz, M., & Marques, P. (2016). Mandating treatment based on interlock performance: Evidence for effectiveness. Alcoholism: Clinical and Experimental Research, 40,  1953-1960.

      This study examined the integration of alcohol use disorder (AUD) treatment with the interlock program.

      Evaluation Methodology

      Design:

      Recruitment: The study used records from the Florida Department of Highway Safety and Motor Vehicles (DHSMV) to identify DUI offenders who installed interlocks between October 2008, when legislation was passed that both required an alcohol use disorder treatment program for offenders with 3 or more interlock violations and that lowered the threshold arrest blood-alcohol content for mandating first offenders to install the interlock, and December 30, 2012.

      Assignment: The study drew from 27,736 records of offenders who installed interlocks after October 2008. From those records, 879 had 3 or more violations – of which 640 had received notices ordering them to enter treatment and were thus included as the treatment group. The study drew a comparison group from 4,642 offenders with 1 or 2 violations who, while not mandated to treatment, had completed their time on the interlock and had their interlock removed and their licenses fully restored. The comparison group (n=806) were selected based on a principal component analysis for 2 underlying factors of recidivism risk: 1) prior record, as measured by the number of prior offenses, arrest blood-alcohol content, interlock sentence length, and the probability of detection if driving impaired; and 2) interlock record, which represented drinking and driving behavior while on the interlock. These two factors, which were correlated at r = 0.132 (so not completely orthogonal), along with “time on the interlock,” were included as the three variables used in a propensity score model to match treatment and comparison groups.

      Attrition: The study does not discuss attrition. Since it uses state records for recidivism, however, there is presumably no attrition.

      Sample: Across the entire sample, the majority of participants were male (80.3%) and white (75.4%). In terms of age, 3.7% were under the age of 25, 21.3% were 25-34 years old, 28.1% were between the ages of 35-44, 30.8% were 45-54, 9.4% were over age 55, and age was unknown for 6.7% of the sample. Approximately half of the sample had 3 or more prior offenses (52.3%) and on average, participants had 2.26 lockouts on interlock (log transformed).

      Measures: The outcome was recidivism after the removal of the interlock, as measured by: 1) DUI arrest or conviction; or 2) indication that an arrest had occurred but had not resulted in a conviction. Recidivism data were supplied by the DHSMV interlock record system, which was matched each month with the state driver record system to update DUI offense data.

      Analysis: The study performed a survival analysis of DUI recidivism using Cox regression proportional hazards model. Those who recidivated more than once were allowed to accumulate multiple “death” dates and could be counted again. Covariates were selected using stepwise regression. Significant predictors included age, time on interlock, and prior record (as defined by a principal component analysis, or PCA). The Interlock Record PCA component was on the cusp of significance and was therefore also included as a covariate.

      In addition, a Heckman selectivity analysis was conducted to assess selection bias (since offenders in the treated group each committed a third violation, whereas the those in the control group committed 1-2 violations). Results were not significant, which the authors concluded demonstrated no bias. To be conservative, the authors tested the sensitivity of the Cox regression recidivism analysis to the inclusion versus exclusion of the Heckman estimate and found that its inclusion did not meaningfully alter the results or significantly affect any of the other predictors in the model.

      Intent-to-Treat: All subjects were included in the analysis.

      Outcomes

      Implementation Fidelity: Participants were mandated to enter treatment so a high level of fidelity is assumed. The study did not address the possibility that incidents of intoxicated driving may not be captured.

      Baseline Equivalence: The authors report the groups were highly similar in demographic characteristics at baseline (pp. 1956). It is assumed that baseline equivalence was established on the variables used in the propensity score model (time on interlock and the two factors produced through the principal components analysis – prior record and interlock record). Still, the treated group had 3 or more interlock system violations (2 lockouts within a 4-hour period while attempting to start their vehicle with a BAC > .05) while the control group had 1-2 violations. In addition, the treated group had a slightly higher but not statistically significant ( = 0.327) composite risk score based on combining the prior record and interlock record recidivism scores produced through the principal components analysis.

      Differential Attrition: As records for all subjects were searched, there appears to be no attrition.

      Posttest: The study found that the proportional hazard of recidivism was 32% lower for the treatment group (mandated to alcohol use disorder treatment) than the matched-risk comparison group.

      Long-Term: The study examined records for a total of 1.5 years after the interlock de-install date, but does not report long-term and short-term results separately and exposure and follow-up time varied in the sample.

      Roth, R., Voas, R., & Marques, P. (2007). Mandating interlocks for fully revoked offenders: The New Mexico experience. Traffic Injury Prevention, 8 (1), 20-25.

      Evaluation Methodology

      Design:

      Recruitment: Between July 1999 and January 2003, New Mexico had a law that made ignition interlocks an optional judicial sanction for second and third-time DUI offenders and another law that required a 1-year hard suspension for second-time DUI offenders. A judicial requirement to install an interlock did not affect the suspension status of the offender. Consequently, 95% of the study’s subjects were revoked at the time they installed interlocks. A total of 12,991 subjects were recruited for the study.

      Assignment: Data for offenders who installed interlocks between July 1999 and December 2002 (treatment group, n = 437) were matched with DWI arrest and conviction records in the New Mexico Motor Behavioral Division DWI Citation Tracking System, which contains records of every driver arrested for a DWI offense in New Mexico (control group, n = 12,554). Propensity score matching was used to identify this comparison group from the 20,949 noninterlock multiple offenders convicted between July 1999 and December 2002. Four variables were dichotomized and included in the propensity score model: age (< 30 and > 30), blood alcohol content (BAC < .16 and > .16 or refused), gender (male and female), and priors (2 and >3). A 16-element matrix was constructed into which both treatment and control cases were distributed, and comparison cases were randomly selected from each of the 16 cells in the matrix. This process maximized the inclusion of available offenders who did not install interlocks and yielded a comparison group that had the exact same proportion in each of the 16 matching categories as in the interlock group.

      Attrition: Participants were followed for a 4-year period beginning 70 days after conviction (the average point at which offenders installed interlocks) that included both installation and post-installation conditions for the treatment group compared to the control group. Since administrative data were used to measure outcomes, it is assumed there was no attrition and all records were used in the analysis.

      Sample: The treatment and control groups were each composed of 84% males, 73% of whom were aged 31 or older. At baseline, each group had 51.6% second offenders and 49.4% third or more frequent offenders, and 66.8% of the members of each group had arrest blood alcohol contents of .16 or higher or had refused the breath test at the time of their key arrest.

      Measures: DWI arrest and conviction data for all study participants were received from the Motor Vehicle Departments’ Citation Tracking System in New Mexico. While complete within state, the figures missed recidivism outside of New Mexico.

      Analysis: Survival rates by condition were compared using Cox regression with the same bivariate variables used to construct the propensity score model (blood alcohol content, age, gender, and priors). Recidivism events included rearrest for DWI, and three analyses were conducted: (1) while the interlock was on the offenders’ vehicles (approximately 322 days), (2) during the interlock removal, or for approximately 3 years to the end of the study period, and (3) for the sum of periods 1 and 2 which encompassed four years from the starting point as the first analysis.

      Because the average time between conviction and installation for the treatment group was 70 days (see implementation fidelity), the index time for the comparison group was shifted to 70 days after conviction for the first and third analyses, thus ensuring the beginning of the exposure period was the same for both groups. For the second analysis, an additional 281-day shift of the index time for the control group was included to match the average interlock removal time of those in the treatment group (once again, see implementation fidelity). Those who reoffended before the index time were excluded from each analysis.

      Intent-to-Treat: It seems that all participants assigned to a condition were included in the analysis.

      Outcomes

      Implementation Fidelity: Of the 437 interlock cases, 95% (n = 415) removed the units before the end of the study period. Meanwhile, 87% (n = 378) of the offenders in the treatment group installed interlocks within one year of conviction. The mean time between conviction and installation was 0.19 years or 70 days, and the mean time of installation period was .77 years or 281 days. The authors note that 41 records showed installation durations longer than 400 days, likely composed of offenders for whom the probation department may have extended the requirement or who voluntarily kept the interlocks installed.

      Baseline Equivalence: Propensity score matching (PSM) was used to match the control group with the treatment group, and the effort to distribute equal proportions of both conditions in the 16 cross-classified cells indicates that the groups were equivalent on the four dichotomous measures used in the PSM model (age, gender, blood alcohol content and prior DWIs). The variables in the model were the only data available for the study. Because the study used a survival analysis, testing baseline equivalence between groups on recidivism outcomes was not possible.

      Differential Attrition: With all subjects assumed to be followed, there was no attrition.

      Posttest: The control group showed a significantly higher recidivism rate than the treatment group from baseline to follow-up. That is, during the time when the interlocks were installed, the recidivism rate of the treatment group was estimated to be 0.34 of the comparison group’s (p = .0004) with a relatively large 95% confidence interval (0.19 – 0.62). There were no significant differences between groups during the period after the interlocks were removed (and after excluding subjects who reoffended earlier). Specifically, the analysis during the post-interlock period showed the recidivism ratio was 0.96 but was not statistically significant (p = .74).

      Long-Term: The analysis covering a 4-year period following the same baseline as was used in the first analysis revealed the rearrest rate of the treatment group was 0.78 of that for the control group (p = .02). Thus, long-term results show that the treatment group still had lower total recidivism than the control group.

      Willis, C., Lybrand, S., & Bellamy, N. (2009). Alcohol Ignition Interlock Programmes for reducing drink driving recidivism (Review). The Cochrane Library, 1, 1-26.

      Evaluation Methodology

      Design:

      Recruitment: This meta-analysis included only RCTs and other controlled trials that examined samples of persons previously convicted of drunk driving. The study searched eight databases, numerous web pages, and several conference proceedings, and also asked authors of published works for completed and unpublished studies. Two reviewers independently rated the studies. Fifteen met the selection criteria, but one was dropped for overlap with another study, leaving 14 for the review.

      Assignment: Of the 14 studies, one was an RCT and 13 were controlled trials. Of the controlled trials, one was an effectiveness study and the others were efficacy studies. Of the efficacy studies, one was a before-and-after study, two were retrospective record reviews, and six were longitudinal studies. The effectiveness study was a before- and after-design. The three remaining trials were ongoing, and had yet to be reported.

      Three of the controlled trials were administered through the courts. Without randomization, judges may have chosen offenders with certain characteristics for the intervention group. Some programs were ’semi-mandatory,’ giving rise to self-selection bias, whereby those who chose to participate in the program may have been more motivated to succeed.

      Four of the studies had a control group in which people were legally allowed to drive. The other studies had control groups that were still suspended, potentially being more cautious and driving less miles, which would therefore limit the exposure to being caught.

      Sample:

      The studies examined samples from multiple cities and states in the U.S. plus samples from Canada, Australia, and Sweden. Most of the studies included only the most motivated subjects, as participation rates were low.

      Measures:

      Recidivism for drunk-driving offenses was measured while the interlock device was installed on the offender’s vehicle, after the interlock device had been removed from the offender’s vehicle, and for the full study period.

      Analysis:

      The primary analysis was based, for RCTs, on meta-analytic methods. Relative risk with random effects was calculated using RevMan 4.2.


      Results from the non-RCTs were considered in the discussion but were not part of the meta-analysis, due to differences in methodology and potential biases.

      Outcomes

      For the one RCT (see Study 2 above), recidivism was lower in the intervention group while the device was still installed in the vehicle; relative risk 0.36 (95% confidence interval 0.21 to 0.63). The benefit disappeared once the device was removed; relative risk 1.33 (95% confidence interval 0.72 to 2.46).

      In all 13 non-RCTs, interlock participants again had lower recidivism than the controls. In nine of the trials, the difference between the groups would be regarded as statistically significant. For repeat offenders, the evidence was stronger, with six of the eight studies showing that the offenders had significantly lower recidivism rates while the device was installed in the vehicle. Once again, however, the favorable result did not extend to the time period after the interlock was removed.

      The two controlled trials reporting an overall effect for the full time period both showed that the interlock did not reduce recidivism in the drunk driving population.

      Summary. The review found evidence that the ignition interlock had the ability to curb drinking and driving when it was installed in a vehicle but that the benefit disappeared once the device was removed. The lack of long-term results and the low rate of participation in the program made overall effectiveness of the device questionable from a traffic safety point of view.

      Even these conclusions are limited. Only one study was an RCT, but it included only more motivated repeat drunk drivers who had been successful enough to apply for re-licensing. The low participation rates in the non-randomized trials also weakened the findings. The program has the potential to reduce the frequency of DUI and may favorably influence the burden and costs of alcohol-related road traffic accidents, but the quality of the evidence needs to be improved.